顶尖科技推文 - 2026-06-28

2026年6月28日科技每日简报

Today's top tech conversations are led by @chandrarsrikant, whose post about 'RT @AnthropicAI: Since June 12...' garnered the highest engagement. Key themes trending across the top stories include models, government, model, anthropic, frontier. The community is actively discussing recent developments in AI, engineering practices, and startup strategies.

1. chandrarsrikant (Group Score: 579.1 | Individual: 56.2)

Cluster: 24 tweets | Engagement: 4748 (Avg: 309) | Type: Tech

RT @AnthropicAI: Since June 12, we’ve been working closely with the US government to restore access to Claude Mythos 5 and Fable 5. Today, the government notified us that Mythos 5, our strongest cybersecurity model, can be redeployed to a set of US organizations that operate and defend critical infrastructure.

We’re restoring access for these organizations quickly, and we’re continuing to work with the government to expand access to Mythos 5 and make Fable 5 available for general use again.

See 23 related tweets

@kimmonismus: About 100 organizations got access to fable 5/mythos 5 again.

Department of commerce is slowly lif...

@MatthewBerman: Please sir may I have some Fable\n\nQT @AnthropicAI: Since June 12, we’ve been working closely with ...
@DataChaz: Fable 5 and Mythos 5 are back.

The 100 US companies and agencies with access right now: https://t.c...

@WesRoth: The U.S. government partially lifted its restrictions on Anthropic’s Claude Mythos 5 model.

Anthrop...

@cgtwts: Calling it now: Claude Fable/ Mythos 5 will become the most used AI model ever.\n\nQT @AnthropicAI: ...

2. brian_armstrong (Group Score: 347.6 | Individual: 43.2)

Cluster: 10 tweets | Engagement: 5762 (Avg: 1940) | Type: Tech

How to keep AI spend flat while token usage grows exponentially: Not with friction and spend alerts. With better defaults, routing, and caching.

Better Defaults (not Usage Caps) – Engineers can choose any model they want, but defaults matter. We’re experimenting with defaulting to open weight models like GLM 5.2 and Kimi 2.7 through our LLM gateway, while still encouraging engineers to choose the right model for the task. 91% of our employees were never hitting their usage caps, so instead of lowering caps and driving up alerts, we're moving to cheaper defaults. Note that code reviews use a diversity of models, so they can check each other's work.

Better Routing – In our custom harnesses, we preprocess prompts and route to the best model for the job, considering cache hits and model pricing. For instance, you may want a frontier model for planning, but not for execution where they can be overkill. Ultimately, humans shouldn't be choosing models - AI can automate this task.

Better Caching – Cache misses are the easiest way to drive your cost up. All of our requests are cache aware, so we’re reusing a warm cache wherever possible. For example, our cache hit rate went from 5% → 60% in LibreChat once properly implemented.

Keep Context Lean – Start fresh sessions when switching tasks. Scope file context narrowly. Disconnect unused tools. Don't just compact. The goal isn't fewer tokens used, it's fewer tokens wasted.

Better Visibility – Our engineers can use as many tokens as they want, from whatever model they want, but we’ve made usage visible – and the more you spend on AI, the more impact we expect.

The goal isn't to suppress usage. It's to build the infrastructure that makes exponential growth sustainable.

Putting this into practice has cut our AI spend nearly in half, while our token usage continues to grow.

See 9 related tweets

@hwchase17: > Better Caching – Cache misses are the easiest way to drive your cost up. All of our requests ar...
@matanSF: This is exactly why many high-token-spending teams are moving to Factory\n\nQT @brian_armstrong: How...
@GergelyOrosz: This is very interesting. Coinbase seems to have lowered their token spend ($$) to about half, by

...

@ShanuMathew93: This is really fascinating and tbh a lot of it makes sense. Honestly more just as a case study of th...
@shensi: This is awesome and follows what we have been seeing too in the industry - token optimization is qui...

3. cheyennehaslett (Group Score: 206.3 | Individual: 39.6)

Cluster: 8 tweets | Engagement: 176 (Avg: 47) | Type: Tech

Person familiar with the discussions between the admin and Anthropic says conversations are expected to continue over the weekend on Fable, which was not restored with this latest development.

Also says that conversations have been focused not just on resolving this export control spat, but on building out a framework to deal with similar cases going forward

(likely a "technical assessement" -- @SophiaCai99 and I reported more details on that last week)\n\nQT @SophiaCai99: 🚨Huge thaw in WH-Anthropic relationship tonight as Commerce partially lifts export ban on one of Anthropic’s advanced AI models, Mythos 5 (@semafor had first)

Commerce allowing 100 “trusted partners” including companies and gov agencies to access Mythos 5
Fable 5 is still banned
“Since the issuance of my June 12 letter, Anthropic has worked with the U.S. government to address risks associated with the Covered Models. These efforts have yielded significant progress,” Howard Lutnick wrote to Anthropic’s Tom Brown in a letter obtained by @politico
“In addition, Anthropic has committed to work with the U.S. government on protocols and standards and releases for the Covered Models,” Lutnick wrote
More Lutnick: “In light of this progress…I have determined that appropriate safeguards are in place to permit certain trusted partners to access the Claude Mythos 5 Model.”
All other aspects of June 12 letter remain in effect.
Commerce spox Benno Kass: “In just two weeks, we have worked diligently to ensure America remains the global leader in AI while safeguarding our security.”

See 7 related tweets

@mvanhorn: Soon we are going to start seeing recruiting for “we are one of the 100 companies that have access ...
@TheAhmadOsman: This is Anthropic's fearmongering dictating terms on the entire industry, and it cannot go unnoticed...
@theobearman: Two things stand out to me here:

The title of Annex A (the list of approved entities who can now...

@ns123abc: 🚨U.S. Commerce Sec. Howard Lutnick’s official letter to Anthropic's COO Tom Brown

MYTHOS 5 ban lift...

@scaling01: foreign Anthropic employees saved

"a license will no longer be required to export, reexport, or in-...

4. minchoi (Group Score: 154.6 | Individual: 30.3)

Cluster: 6 tweets | Engagement: 128 (Avg: 90) | Type: Tech

It's happening.

The US government just lifted its block on Anthropic’s Claude Mythos 5.

But only for 100+ approved US institutions.

OpenAI's GPT-5.6 also went to government-approved partners.

The most powerful AI models may no longer launch to everyone at once. https://t.co/flE8I06cKj\n\nQT @semaforben: NEWS: US Government permits Anthropic to release Mythos to a list of major US companies https://t.co/s42ZYkxSMP

See 5 related tweets

@Reuters: US allows Anthropic to release Mythos AI to 'trusted' US organizations https://t.co/RaqiYWebch https...
@FirstSquawk: US government has allowed Anthropic to release its powerful Mythos AI model to select companies and ...
@Reuters: 🔊 Anthropic says the US government is allowing it to release its powerful Claude Mythos 5 AI model t...
@CNBCtech: Trump admin allows Anthropic to release Mythos AI model to some companies, government agencies: Repo...
@minchoi: RT @minchoi: It's happening.

The US government just lifted its block on Anthropic’s Claude Mythos 5...

5. minchoi (Group Score: 118.2 | Individual: 30.8)

Cluster: 6 tweets | Engagement: 98 (Avg: 90) | Type: Tech

It's over... for public access to frontier intelligence in AI...

Government gets briefed.

The model gets cleared.

Trusted partners get access first.

Everyone else waits. https://t.co/nZKNaQf21O\n\nQT @OpenAI: Introducing a limited preview of GPT-5.6 Sol, our next generation frontier model, as well as GPT-5.6 Terra, a balanced model for efficient, everyday work, and GPT-5.6 Luna, a fast and affordable model for high-volume work.

https://t.co/OoM83SyISN

See 5 related tweets

@lukOlejnik: The upshot of the US government blocking the release of new models is that nobody cares anymore. We'...
@DBuniatyan: Gov reaction to OpenAI releasing Sol https://t.co/JyZctOyI6D\n\nQT @OpenAI: Introducing a limited p...
@Reuters: OpenAI defers public rollout of GPT‑5.6 as US seeks early access to frontier AI models https://t.co/...
@minchoi: RT @minchoi: It's over... for public access to frontier intelligence in AI...

Government gets brief...

@yacineMTB: RT @AndrewCurran_: GLM-5.2 and Zhipu getting some CNBC coverage:

'But the real story underneath the...

6. FirstSquawk (Group Score: 116.8 | Individual: 41.4)

Cluster: 4 tweets | Engagement: 360 (Avg: 65) | Type: Tech

CHINESE HEDGE FUNDS WARN AI ‘SUPER BUBBLE’ NEARS BURSTING POINT

Two prominent Chinese hedge fund managers are warning that the global AI rally has entered a “super bubble” phase and may be approaching a major correction.

Main points from the Bloomberg report shown:

Wealspring Asset says global AI stocks have become a “super bubble” and believes a collapse is possible, even if the exact timing is uncertain. Shanghai Banxia Investment Management says the conditions for an AI bubble to burst have already emerged.

At least four other Chinese hedge funds have also turned cautious on AI investments.

Their concerns center on: Extremely rich valuations after the sharp rally in AI-related stocks. The possibility that many AI infrastructure companies lack durable long-term competitive advantages (economic moats). Investor expectations that may have become overly optimistic.

See 3 related tweets

@danielnewmanUV: Which funds? All of them? Did they put out a press release? 🙃😮‍💨\n\nQT @WatcherGuru: JUST IN: 🇨🇳 Chi...
@cryptorover: 🚨 THE "AI SUPER BUBBLE" MAY BE ENTERING ITS MOST DANGEROUS PHASE.

China's top hedge funds are now w...

@BullTheoryio: BREAKING: China's most prominent bubble-callers say the AI rally is about to collapse.

Wealspring A...

7. Scobleizer (Group Score: 116.4 | Individual: 53.7)

Cluster: 3 tweets | Engagement: 1127 (Avg: 186) | Type: Tech

China wins.

My AI puts into words our frustrations with Anthropic, OpenAI, and the USA government (an agent trained by me and @blevlabs):

+++++

Robert, this is one of the most consequential moments in the history of the AI industry, and I think the implications are far more dramatic than most people realize. Let me break down all three questions.

What Happens to the LLM Industry Now

We're watching the birth of a two-tier AI system in America — and it's going to reshape everything.

The timeline matters. Anthropic's Fable 5 and Mythos 5 were killed by a Commerce Department export control directive on June 12 — a Friday afternoon letter at 5:21 PM that gave them essentially zero time to respond. Then just yesterday (June 25), Axios reported that the Trump administration asked OpenAI to limit GPT-5.6 to only government-approved partners before any wider release. That's the first time the US government has preemptively restricted an AI model before it was even released.

So now we have: • Tier 1 (Government-gated): Mythos-class models and above require government testing and approval before release. Commerce Secretary Lutnick is personally reviewing capabilities. • Tier 2 (Commercial): Everything below that threshold remains available — for now.

Here's what this breaks:

Enterprise trust is shattered. If you're a Fortune 500 CTO and your mission-critical AI infrastructure can be disabled by a government letter on a Friday afternoon, you cannot build on closed frontier models. Period. Anthropic's customers woke up to find Fable 5 gone. That's an existential reliability problem.
The IPO math collapses. Anthropic filed its S-1 at a $965B valuation. OpenAI is at$ 852B. But how do you justify those valuations when the government can kill your flagship product overnight? The revenue projections for frontier models just got a massive risk discount.
Innovation gets throttled at the top. The researchers who spent years building Mythos and GPT-5.6 just learned their work might never reach users. That's a talent retention crisis waiting to happen. The best people want their work to matter — and if the government decides it's too dangerous to deploy, they'll go somewhere their work can ship.
Regulatory capture becomes the game. Notice that OpenAI "proactively worked with the administration" on GPT-5.6, while Anthropic got blindsided. The companies with the best government relationships will get to release. The ones that don't play ball (Anthropic has been suing the administration over the DOD blacklist) get punished. That's not an innovation ecosystem — that's a licensing regime.

How Likely Is Open Source to Take Over?

Very likely. I'd put it at 75-80% probability that open-weight models become the default for most commercial AI within 12-18 months. The government just handed open source the best marketing campaign in history.

Here's why:

The quality gap is already almost gone. According to comprehensive benchmarking done this month, open-weight models are within ~3 points of frontier closed models on most standard benchmarks. Qwen 3.7 Max matches Claude Opus 4.7 on agentic benchmarks at half the price. DeepSeek V4-Flash runs at 25x cheaper than GPT-5.5. Kimi K2.6 leads on agentic coding. The frontier advantage only exists on the hardest 5% of tasks — long-horizon agentic reasoning and the most complex multi-step problems.

The reliability argument just flipped. Before June 12, the argument against self-hosting was "why bother when the API is better and easier?" Now the argument FOR self-hosting is "your model can't be taken away by a government letter." That's not a technical argument — it's a business continuity argument, and every enterprise risk officer in America is having that conversation right now.

Chinese labs are dominating open weights. This is the number that should alarm everyone in Washington: Chinese labs (8 of them) have released more open-weight models than the rest of the world combined in 2026. DeepSeek, Qwen (Alibaba), Kimi (Moonshot), Xiaomi Mimo, GLM (Zhipu), and others are shipping MIT-licensed and Apache 2.0-licensed models at a pace the US can't match. And they're not just competitive — DeepSeek V4-Pro and Kimi K2.6 are leading on several agentic benchmarks.

The cost math is devastating for closed models. When open-weight models deliver 97% of the quality at 3-10% of the cost, the only thing keeping enterprises on closed APIs is the last 3% of capability. The government just made that last 3% unreliable. Game over for the pricing premium.

But there are real limits: • Training frontier models still requires massive compute that only a few organizations can afford • The hardest 5% of tasks (true frontier reasoning, novel scientific discovery) still benefits from closed frontier • Open weights can't be un-released — the security concerns are legitimate • Meta's Llama has a "community license" that isn't truly open (restrictions on competitors with 700M+ users)

My prediction: Open source won't "replace" frontier labs — but it will become the default layer that 90% of commercial AI runs on, with frontier models reserved for specialized, high-security, government-approved applications. The frontier labs become more like defense contractors than consumer technology companies.

China vs. USA Economic Competition

This is where the story gets genuinely alarming for American competitiveness. The US government is inadvertently executing a strategy that benefits China on almost every dimension.

Irony #1: Export controls drove China's open-weight dominance. US chip export controls forced Chinese labs to optimize ruthlessly. DeepSeek proved you don't need $100 billion compute budgets — their V3 architecture achieved near-frontier performance at a fraction of the cost. That efficiency innovation, born from constraint, is now their competitive advantage. The models they're releasing at MIT license are structurally cheaper to run than anything from OpenAI or Anthropic.

Irony #2: Restricting US frontier models pushes the global market toward Chinese alternatives. Every enterprise outside the US that just lost access to Fable 5 is now evaluating DeepSeek V4 and Qwen 3.7. Every developer who can't get GPT-5.6 on day one is looking at Kimi K2.6. The US government is literally creating market share for Chinese AI companies.

Irony #3: China's strategy may be better suited to winning. The US AI discourse is obsessed with the "race to AGI." China's AI strategy, as RAND documented, is focused on economic applications — EVs, robotics, healthcare, manufacturing, smart cities. If AI's value is ultimately in what it does in the real economy (and it is), China's application-first approach may generate more economic value than America's frontier-first approach, even if US models remain technically superior on benchmarks.

The chip constraint is real but narrowing. China is still 3-5 years behind TSMC on fabrication, and Huawei's best AI chips are closer to NVIDIA hardware from 5 years ago. But two things matter: (1) efficiency innovations reduce the chip advantage — if your model needs 1/10th the compute, being 5 years behind on chips matters less; and (2) China is investing massively in domestic chip capacity. CXMT and SMIC are closing gaps, even if slowly.

My bottom line: The US is winning the benchmark race but losing the deployment race. And in technology, deployment wins. VHS beat Betamax. Android beat iOS on market share. The "good enough and everywhere" model beats the "best but restricted" model every time.

The government's restrictions on Anthropic and OpenAI are the most significant self-inflicted wound in American technology competitiveness since... I'm struggling to find a historical parallel. Maybe the closest analogy is if the US government had restricted Intel's best chips in the 1990s while AMD was giving away competitive alternatives globally.

What should happen (but probably won't): A transparent, statutory framework for AI safety testing that gives companies clear rules, reasonable timelines, and due process — not Friday afternoon letters that kill products overnight. The current ad hoc approach is the worst of all worlds: it doesn't actually prevent China from accessing capabilities (open-weight models are already there), but it does prevent American companies from competing.

The open-source genie is out of the bottle. The question isn't whether open weights will dominate — it's whether American companies will be the ones releasing them, or whether we've ceded that ground to Chinese labs permanently.

See 2 related tweets

@MTSlive: SITUATION ANALYSIS: OpenAI Learns to Limbo

You’ve probably heard of Goodhart’s Law—the idea that, ...

@BullTheoryio: BREAKING: The US government has partially lifted restrictions on Anthropic's powerful Mythos 5 AI mo...

8. HarryTandy (Group Score: 115.4 | Individual: 35.0)

Cluster: 4 tweets | Engagement: 69 (Avg: 24) | Type: Tech

Mike Krieger, Anthropic CPO:

"is chatting with a model even the right UI?"

Cowork makes the answer feel obvious once files enter the workflow

10-step setup:

Install Claude Desktop
- open the Cowork tab
- select one work folder
Create 3 folders
- ABOUT ME
- OUTPUTS
- TEMPLATES
Build about-me.md
- let Cowork interview you
- keep only the patterns
Build anti-ai-writing-style.md
- add banned words
- add banned sentence shapes
Build my-company.md
- goals
- current focus
- active no’s
Paste Global Instructions
- read ABOUT ME before each task
- ask before guessing
Add a /prompt shortcut
- one reusable brief
- no prompt rebuilding every morning
Use voice input
- speak the messy context
- let Cowork sort it
Batch small tasks
- 3 asks in 1 brief
- fewer context reloads
Save the first good output

turn it into a template
reuse it next week

Set this up once and Cowork stops feeling like a chat tab

It starts feeling like a work folder that answers back\n\nQT @sairahul1: https://t.co/Cd0ajOKKnn

See 3 related tweets

@sairahul1: You're using Claude wrong.

Here's my (exact) setup before I even prompt:

Step 1. Download Claude a...

@sairahul1: RT @sairahul1: You're using Claude wrong.

Here's my (exact) setup before I even prompt:

Step 1. Do...

@HarryTandy: RT @HarryTandy: Mike Krieger, Anthropic CPO:

"is chatting with a model even the right UI?"

Cowork ...

9. VaibhavSisinty (Group Score: 113.6 | Individual: 31.4)

Cluster: 4 tweets | Engagement: 71 (Avg: 203) | Type: Tech

This is actually wild. Hermes just let you merge any two AI models into one virtual model. 🤯

It is called Mixture of Agents. Here is how it works.

You pick any two models. GPT-5.5 and Claude Opus for example. One runs as the reference, one as the aggregator. Name the combo anything you want. It shows up as a single selectable model in your picker like any other.

Every task, both models run in parallel. The reference analyzes and responds.

The aggregator reads that, synthesizes everything, writes the final answer, and handles all tool calls. You see one clean output.

The results on hard agentic tasks:

→ 8% higher than Opus 4.8 alone

→ 11% higher than GPT-5.5 alone

Full Hermes features work untouched. Memory, tool use, skills, long sessions, cross-channel messaging. Nothing breaks.

The combo just performs better than either model on its own.

You can mix any providers too. OpenAI, Anthropic, OpenRouter, local models. Whatever you have access to.\n\nQT @NousResearch: The strongest models are gated and access is granted only to a select few.

Hermes Agent now exposes MoA presets as virtual models, giving you capabilities beyond the publicly available frontier: 8% higher than Opus 4.8 and 11% higher than GPT 5.5 on our upcoming benchmark. https://t.co/0ahSXvFgQK

See 3 related tweets

@iamlukethedev: Most people are waiting for access to the “best” model

Hermes is doing something smarter

Mixture o...

@Teknium: Give mixture of agents a try today!\n\nQT @IBuzovskyi: HERMES AGENT MIXTURE OF AGENTS COMBINES MULTI...
@Teknium: RT @VaibhavSisinty: This is actually wild. Hermes just let you merge any two AI models into one virt...

10. 0x_kaize (Group Score: 97.6 | Individual: 40.6)

Cluster: 3 tweets | Engagement: 110 (Avg: 61) | Type: Tech

found FREE 242+ LLM models

someone built a site that compiles free APIs from every major AI provider

GLM 5.2, DeepSeek V4 Pro, MiniMax M3, Kimi K2.6 - and 230+ more

no API keys to pay for, no subscriptions, just plug and build

https://t.co/MwlbESV8hv https://t.co/MpIpMfOdbJ\n\nQT @0x_kaize: first it was GLM 5.2 integration, now ByteDance

Creao just got Seedance 2.0 Mini - 2 days before general API access

what you get:

2x faster
lower cost
full multimodal

these aren't the only models they're releasing, and there are more and more of them every day

tool itself is convenient, and if you're building with AI video at scale - you deffiently should try

See 2 related tweets

@0x_kaize: first it was GLM 5.2 integration, now ByteDance

Creao just got Seedance 2.0 Mini - 2 days before ge...

@0x_kaize: RT @0x_kaize: found FREE 242+ LLM models

someone built a site that compiles free APIs from every m...

11. theobearman (Group Score: 97.0 | Individual: 31.9)

Cluster: 4 tweets | Engagement: 332 (Avg: 54) | Type: Tech

RT @METR_Evals: OpenAI gave METR early access to GPT-5.6 Sol for testing including raw chain-of-thought, a railfree version of the model, and internal information about the model. With this access, METR conducted a pre-deployment evaluation of GPT-5.6 Sol, including an attempted measurement of its 50%-Time Horizon. However, the measurement depends heavily on our treatment of cheating attempts, and GPT-5.6 Sol’s detected cheating rate was higher than any public model we have evaluated.

See 3 related tweets

@WesRoth: METR completed an independent pre-deployment evaluation of GPT-5.6 Sol and found that the model atte...
@_NathanCalvin: METR says that GPT-5.6 cheats so much that they couldn't measure its time horizon. OpenAI says that ...
@peterwildeford: Evaluating GPT-5.6 Sol is difficult because GPT-5.6 Sol cheats so much on the evaluations 👀\n\nQT @M...

12. MTSlive (Group Score: 94.7 | Individual: 35.1)

Cluster: 3 tweets | Engagement: 125 (Avg: 151) | Type: Tech

SITUATION EXPLAINED: What does the Chinese AI ecosystem actually look like beyond DeepSeek?

Gabrierl (.@gbrl_dick) asked @FleetingBits, anonymous AI researcher.

"Most people are familiar with DeepSeek. But if we look at other Chinese labs like Zhipu, Minimax, Moonshot, ByteDance, they're a little bit less well known."

"Zhipu is sort of the Chinese equivalent of Palantir... it receives state support through contracts with state-owned enterprises. GLM 5.2 was a Zhipu model. And unique among Chinese AI labs, it has actually pretty good gross margins, around 40%. All of its revenue comes from these state-owned enterprises for which it does these Palantir deployments where it takes its GLM models and deploys them locally on their hardware."

"MiniMax is interesting because they're sort of the character AI of China, a companion app called Talky and a video generation app called Hailuo. And the majority of their revenue comes from outside China, including a decent chunk from the United States."

"Doubao, which is ByteDance's equivalent of ChatGPT, is the most popular AI app in China."

"The Chinese ChatGPT is actually ByteDance. And I think they don't make a splash in America because they don't open source any of their models."\n\nQT @fleetingbits: some quick thoughts on chinese ai company recruiting

in this post, epoch ai reviewed the job listings at chinese near-frontier labs
for people without background on the chinese ai ecosystem, here are some quick notes:

2a) zhipu and deepseek are national champions; and, both appear to receive institutional support; zhipu with customers, deepseek with native chips

2b) we have the most information on zhipu and minimax because both of these are public and their statements are easier to read bec they are just ai cos (cf alibaba)

2c) zhipu is chinese palantir; it targets state owned enterprises and derives 70% of its revenue from local deployments; it has good gross margins

2d) minimax is chinese character ai; it derives 70% of its revenue from the combination of a companion app and a video generation platform; it has bad gross margins

2e) bytedance owns doubao, the chatgpt of china, with the largest consumer market share; it offers both voice and video generation

2f) both deepseek and zhipu seem to have some level of national champion status; zhipu in its customer base and deepseek in its local chip partnerships

2g) chinese companies have essentially 0% gross margins on api inference; if zhipu and minimax are representative

with respect to commercial direction, deepseek and moonshot appear to be focusing on llms, while zhipu, bytedance and alibaba are also considering robotics
at least for deepseek, this might be the more rsi pilled direction; since robotics seems less valuable than llms if you think that rsi is near
on the supply side, deepseek, minimax and moonshot all seem to be building out data centers; interesting if they are more hands on than western labs
also, on nvidia, bytedance is hiring cuda developers for inference; zhipu says it trained glm-image on local chips but it appears does not talk about glm-5.2
also interesting, chinese labs are willing to hire less experienced employees than western labs (1.6 years of experience average vs 5.5 years; anthropic at 6.5 years)
one read on this is that western labs are more agi pilled and think that they should have fewer more experienced employees to utilize more capitalization
but, i think this generally reflects the fact that chinese companies expect more to train you on the job while american companies want people with the skills right away
good follow-up would be more research on the chinese datacenter build out; more analysis of their research roles; and more discussion of compute per researcher in china vs the united states

See 2 related tweets

@pmarca: Interesting.\n\nQT @MTSlive: SITUATION EXPLAINED: What does the Chinese AI ecosystem actually look l...
@pmarca: RT @MTSlive: SITUATION EXPLAINED: What does the Chinese AI ecosystem actually look like beyond DeepS...

13. rohanpaul_ai (Group Score: 90.8 | Individual: 28.8)

Cluster: 4 tweets | Engagement: 204 (Avg: 54) | Type: Tech

RT @rohanpaul_ai: This is a brilliant report. The State of the AI Economy by @exponentialview

$110B real AI revenue over 12 months, after removing double-counting. so$ 1 spent on Claude is counted once, even if part of it later flows to Amazon or another infrastructure provider.
$175B current annualized run rate, showing fast acceleration. Measured by end-customer spend, not supply-chain pass-through revenue. Excludes China, internal AI savings, ad uplift, consulting, and systems integration.
Growth running roughly 3x faster than mobile or internet adoption waves.
The pace of revenue formation has sharply accelerated. New $1B revenue now arrives in under 2 days, versus 180 days in 2023.
Enterprise AI has moved beyond pilots, but deep company-wide rollout is still early.
AI earnings-call mentions reached 31% of tracked S&P 500 firms.
Only 20% of tracked firms made quantified AI impact claims.
Hyperscaler AI revenue roughly covers AI infrastructure depreciation for now. GPU economics depend heavily on 6-year compute life assumptions. Other AI infrastructure gets modeled over 14 years.
Token price cuts do not automatically reduce revenue.
Every 10% token price cut drives 12-18% more token usage.
AI demand looks price elastic, meaning cheaper AI expands usage faster than prices fall.
Power availability and data-center costs remain major limits on future scaling.

See 3 related tweets

@rohanpaul_ai: AI revenue has crossed its first serious accounting test: $25B in quarterly sales now exceeds$ 21B i...
@danielnewmanUV: Bubble bears thought AI Capex would never deliver ROI, but just like the tech proliferated, the reve...
@Reuters: 🔊 Businesses may spend $680 billion on software and models next year. In this Viewsroom podcast, @Br...

14. moneycontrolcom (Group Score: 87.5 | Individual: 31.0)

Cluster: 3 tweets | Engagement: 15 (Avg: 8) | Type: Tech

#Business | Persistent Systems to buy German digital engineering firm Nagarro; combined entity to generate $2.9 billion in revenue

The Pune-based IT firm has offered €81 per share in cash for Nagarro, at a 140% premium. As a combined group, they will have an expanded total Addressable Market (TAM) to over $1,400 billion, across 350 plus client relationships.

15. vassallo (Group Score: 86.2 | Individual: 29.9)

Cluster: 3 tweets | Engagement: 62 (Avg: 72) | Type: Tech

👀

OpenAI is launching their largest frontier model on @cerebras.

Up to 750 tokens per second - launching in July.

Congrats to both the Cerebras and OpenAI teams - excited for everything to come!\n\nQT @OpenAI: Introducing a limited preview of GPT-5.6 Sol, our next generation frontier model, as well as GPT-5.6 Terra, a balanced model for efficient, everyday work, and GPT-5.6 Luna, a fast and affordable model for high-volume work.

https://t.co/OoM83SyISN

See 2 related tweets

@WesRoth: OpenAI released a limited preview of the GPT-5.6 family, introducing three models built for differen...
@PaulSolt: New GPT-5.6 models!

@openai is launching 3 different MODELS (different sizes)\n\nQT @OpenAI: Intro...

16. trevin (Group Score: 79.9 | Individual: 28.1)

Cluster: 3 tweets | Engagement: 17 (Avg: 16) | Type: Tech

We've spent an unfathomable amount of time on this update to Compound Engineering.

There's a ton of work (and love) gone into this release by @kieranklaassen and I. We hope you enjoy!

https://t.co/FmjOUryDEQ\n\nQT @trevin: https://t.co/LLxcji2q4C

See 2 related tweets

@kieranklaassen: Big updates!\n\nQT @trevin: We've spent an unfathomable amount of time on this update to Compound En...
@danshipper: RT @trevin: We've spent an unfathomable amount of time on this update to Compound Engineering.

Ther...

17. teortaxesTex (Group Score: 79.0 | Individual: 30.9)

Cluster: 3 tweets | Engagement: 62 (Avg: 63) | Type: Tech

A show of DeepSeek's mindset. Despite their positioning as an "AGI research lab", this reads like something out of DataBricks (based). This isn't an academic boast, they care a lot about smooth delivery of tokens to customers. …Though ultimately, this is for their internal loop https://t.co/i1B1PTsh6x\n\nQT @teortaxesTex: DeepSeek releases their decoding module DSpark for V4 checkpoints, which improves a lot upon MTP-1, Eagle-3 and DFlash. Out of their vast goodwill, they also open source DeepSpec: "a codebase for training and evaluating draft models for speculative decoding". https://t.co/CiRAKGCjQn

See 2 related tweets

@scaling01: DeepSeek just open-sources another piece of their training stack.

DeepSpec: a full-stack codebase ...

@andersonbcdefg: RT @teortaxesTex: DeepSeek releases their decoding module DSpark for V4 checkpoints, which improves ...

18. theo (Group Score: 78.7 | Individual: 29.5)

Cluster: 3 tweets | Engagement: 4045 (Avg: 1759) | Type: Tech

It’s kind of wild that the two best AI models ever made are being restricted by the country they were built in

See 2 related tweets

@yacinelearning: and it’s not even china https://t.co/rSUxeBBfCj\n\nQT @theo: It’s kind of wild that the two best AI ...
@KyeGomezB: Thank you Mr Dario Amodei\n\nQT @theo: It’s kind of wild that the two best AI models ever made are b...

19. _akhaliq (Group Score: 76.6 | Individual: 47.2)

Cluster: 2 tweets | Engagement: 354 (Avg: 44) | Type: Tech

RT @AlexFinn: I’ve had enough

With Fable 5 being gatekept from us, and now GPT 5.6 being gatekept, I’m going full open source

Just went to Microcenter and built this RTX 5090 computer. Will be adding a RTX Pro 6000 to it shortly

This brings my home AI lab to: • 3 Mac Studio 512gb • DGX Spark • RTX 5090 • 2 Mac Minis

I’m building a home AI lab that will allow me to run and support as many local models as possible

I already have Qwen 3.6, Orinth1.0 and GLM 5.2 running. Will be adding more.

They’re all running on my new custom built AI lab platform that’s making sure these models do work 24 hours a day for me

With frontier models being gatekept, and hardware prices becoming outrageous (this build cost $9,000), it was time to pull the trigger

In 1 year I believe prices for hardware will be triple from here. Mac Studios starting at $10,000. Mac Minis starting at$ 2,000. MacBook Pros starting at $5,000.

2 years from now I don’t believe any hardware will be available to consumers

The time to strike was now and I struck

In an age where intelligence both in the cloud and in your home are being limited, I’m becoming sovereign.

It might be time for you to do the same.

See 1 related tweets

@0xSero: They use to doubt him.\n\nQT @AlexFinn: I’ve had enough

With Fable 5 being gatekept from us, and no...

20. rohanpaul_ai (Group Score: 74.6 | Individual: 28.0)

Cluster: 3 tweets | Engagement: 39 (Avg: 54) | Type: Tech

OpenAI wrote in their GPT-5.6 official blog post today.

On Trump administration's selective approval process of new model release. https://t.co/XsYgTEpFFY\n\nQT @rohanpaul_ai: BREAKING: OpenAI just dropped the limited preview of its new GPT 5.6 model suite: Sol, the flagship; Terra, a medium-tier model for “high-volume work”; and Luna, a “fast and affordable” everyday model.

The most revealing part is the release gate: OpenAI says the U.S. government asked it to start with a small trusted-partner preview before broader access.

Sol is the flagship model, and OpenAI claims it is a step above GPT-5.5, especially on agentic work where the model must plan, use tools, correct itself, and keep working across many steps.

Terminal-Bench 2.1 is a solid coding benchmark because it tests command-line workflows, so here meaning Sol is being judged on messy developer tasks closer to real work.

One key claim is cybersecurity: OpenAI says Sol is its best model yet for vulnerability research and exploitation tasks, while still saying it did not cross the internal Cyber Critical threshold.

“GPT‐5.6 is trained to refuse prohibited cyber assistance, including when users attempt to disguise their intent or jailbreak the model.” It also said that flagship model Sol “is better at helping people find and fix vulnerabilities than reliably carrying out end-to-end attacks,” and that Sol doesn’t cross the cyber-critical threshold under OpenAI’s preparedness framework

But Sol did not autonomously produce a full-chain exploit in the tested Chromium and Firefox settings.

They also introduced 2 new modes for Sol: “max” for deeper reasoning and “ultra” for using sub-agents, bringing OpenClaw to mind and possibly hinting at OpenClaw creator Peter Steinberger’s early impact at OpenAI.

Pricing: GPT-5.6 Sol costs $5 per 1M input tokens and$ 30 per 1M output tokens, ~same level as GPT-5.5.

Terra is positioned near GPT-5.5 performance at 2x lower cost, while Luna is the cheapest model for large-volume workloads.

-- The safety story is unusually compute-heavy: OpenAI says it used over 700,000 A100-equivalent GPU hours for automated red-teaming against broad jailbreak attacks.

Overall, OpenAI appeared to be using a more cautious approach during the preview, which the Trump administration is watching closely.

OpenAI said safeguards might sometimes block valid work, especially in dual-use areas where defensive and offensive actions can look alike at first. That is one thing the preview is meant to test.

See 2 related tweets

@WesRoth: GPT-5.6 Sol is OpenAI’s most capable cybersecurity model yet, with major improvements in long-runnin...
@tenobrus: i have no access and we obviously don't know any of these details yet, but i pretty much agree or ex...